This week we feature articles from Dean Allemang, and Tim Neusesser & Evan Sunwall.

Additional reading comes from Amelia Wattenburger, Google I/O presenters, and Allen Institute staff.

News comes from Tridion, Adobe & Google, MindsDB & Nixtla, and Algolia.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

AI’s Woolf at the door — LLMs and knowledge graphs

The combination of Large Language Models with other sources of data and knowledge has enormous potential for new, and more powerful, applications across domains. Most of the examples we’ve looked at in the last few weeks involve knowledge graphs and LLMs, which seem like particularly promising partners. Dean Allemang reports from last week’s Knowledge Graph Conference that there is a lot of enthusiasm for such a marriage. But Allemang wants more evidence that knowledge graphs are in fact a better match for LLMs than other types of databases. His article is a good place to start for anyone who has the same question. (8 min)

https://medium.com/@dallemang/ais-woolf-at-the-door-llms-and-knowledge-graphs-eecd6289380f

Error-message guidelines

Who among us hasn’t been occasionally baffled by error messages that aren’t the least bit helpful? This is one of the most frustrating user experience failures. Tim Neusesser and Evan Sunwall describe what designers should watch out for, and review good and bad error-message examples (including one they like from ChatGPT – see the first item in the list just below for more on chatbot UIs.) (7 min)

https://www.nngroup.com/articles/error-message-guidelines/

More Reading

• Why chatbots are not the future of LLM UI, via Amelia Wattenburger
• A good place to start if you weren’t there. There are also links to more detailed presentations. Google I/O ’23 in under 10 minutes, via Google
• Help from the Allen Institute… AI2 OLMo, an open language model made by scientists, for scientists, via AI2 Blog

All Gilbane Advisor issues

---

Content technology news

Tridion unites web content, structured content and headless delivery

Includes Tridion Docs 15, Tridion Sites 10, updated Tridion Dynamic Experience Delivery, and embedded Fonto for structured content authoring.
https://www.rws.com

Adobe and Google integrating Firefly and Bard

Adobe will use the Content Authenticity Initiative’s (CAI) open-source Content Credentials technology for transparency to images generated.
https://www.adobe.com/sensei/generative-ai/firefly.html

MindsDB and Nixtla enhance time-series forecasting

The integration allows developers using MindsDB to build AI forecasting capabilities and anomaly detection in the database without extensive code.
https://mindsdb.com ■‍ https://www.nixtla.io/

Algolia launches AI-powered Algolia NeuralSearch

NeuralSearch uses Large Language Models (LLM) and Algolia’s Neural Hashing for hyper-scale, and constantly learns from user interactions.
https://www.algolia.com/products/neuralsearch/

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature articles from Minqi Jiang, and Scott Brinker.

Additional reading comes from David Weinberger, Mike Masnick, David Pierce, and Kevin Schaul, Szu Yu Chen & Nitasha Tiku.

News comes from Weaviate, Xyleme, Bloomreach, and Replica Analytics.

Time for our own version of spring break. The next issue will be published on May 17.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

Software²: A new generation of AIs that become increasingly general by producing their own training data

“… exploration, as typically studied in reinforcement learning (RL) and supervised learning (SL)—where it appears as some variation of active learning—is primarily designed with a static, predefined dataset or simulator in mind … unsuitable for learning in open-ended domains like the real world, where the set of relevant tasks is unbounded and cannot be modeled as a static, predefined data generator.”

Minqi Jiang and colleagues look ahead to continuous, open-ended, generalized exploration, to get beyond this limitation. (14min)

https://thegradient.pub/software2-a-new-generation-of-ais-that-become-increasingly-general-by-producing-their-own-training-data/

Exploring the 2nd order effects of generative AI in marketing and martech

Scott Brinker is also looking ahead to what’s coming, beyond the immediately obvious, for marketing technology. (10 min)

https://chiefmartec.com/2023/04/exploring-the-2nd-order-effects-of-generative-ai-in-marketing-and-martech/

More Reading

• David Weinberger trys to pin chatGPT down… chatGPT on why it pretends to know things via David Weinberger
• Mike Masnick is encouraged… Bluesky plans decentralized composable moderation via techdirt
• Can ActivityPub save the internet? Who knows? But this is a useful look at the question. via The Verge
• Inside the secret list of websites that make AI like ChatGPT sound smart, and you can see if yours is one of them! A gifted link… via The Washington Post

All Gilbane Advisor issues

---

Content technology news

Xyleme adds Elevate module to CCMS/LCMS platform

The new product adds dashboard analytics, APIs, and rich integration capabilities to the existing component content management system (CCMS)
https://xyleme.com

Weaviate raises $50 million for vector database

The vector database technology addresses the problem of generating, storing, and searching embedding vectors and their corresponding objects.
https://weaviate.io

Bloomreach integrates ChatGPT with CMS

The ChatGPT Text Generator with Bloomreach Content CMS is designed to support the creation and integration of text into e-commerce web pages.
https://www.bloomreach.com/

Replica Analytics unveils Replica Synthesis 3.0

The latest version of trusted synthetic data generation software comes with a new intuitive user interface.
https://replica-analytics.com/

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature articles from Matt Deitke, and Sung Kim.

Additional reading comes from Mike Masnick, Lior Sinclair, Tim Schopf, and Daniel Tunkelang.

News comes from Adobe, Brightspot, Stardog, and Bloomreach.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

Objaverse: A universe of annotated 3D objects

“Recent advancements in computer vision… and natural language processing… were made possible by the availability of massive datasets. However, existing 3D datasets, such as ShapeNet, are limited in scale and diversity, consisting of only 50,000 objects. In response, we on the PRIOR team at AI2 have developed Objaverse, a vast dataset with nearly 1 million annotated 3D objects, complete with descriptive captions, tags, and animations.” 

(2 min) until you start exploring this free resource.

https://blog.allenai.org/objaverse-a-universe-of-annotated-3d-objects-718ef3d61fd6

Chat with document(s) using OpenAI ChatGPT API and text embedding

“Have you ever wondered how these applications are able to chat with documents that are over 100 or even 1,000 pages long, whereas if you try to do the same thing with ChatGPT, you will get an error message…”

Business analyst Sung Kim shares details of how he assembled a solution to chat with documents, PDFs, and books using ChatGPT API and text embedding. (13 min)

https://blog.devgenius.io/chat-with-document-s-using-openai-chatgpt-api-and-text-embedding-6a0ce3dc8bc8

More Reading

• The unintended consequences of internet regulation, via Techdirt
• Overview of the research landscape combining structured and unstructured knowledge in NLP A decade of knowledge graphs in natural language processing via Towards Data Science
• Make your life easier… Create neural network architecture drawings parametrically & export SVG via alphasignal
• Effective query triage via Daniel Tunkelang

All Gilbane Advisor issues

---

Content technology news

Adobe expands Frame.io collaboration platform beyond video

Frame[.]io now includes rich support for photos and PDFs, helping more teams to seamlessly collaborate in a centralized cloud-based workspace.
https://news.adobe.com/news/news-details/2023/Adobe-Expands-Frame.io-Collaboration-Platform-Beyond-Video/default.aspx

Brightspot expands partner program

To offer comprehensive implementation and maintenance options for building digital, web and content experiences using Brightspot CMS.
https://www.brightspot.com

Stardog introduces Stardog 9

The knowledge graph powered semantic layer has new integrations for Azure Synapse, Collibra Data Governance and Databricks.
https://www.stardog.com/blog/introducing-stardog-9/

Bloomreach unveils SMS marketing updates 

The TCPA List Validation feature removes reassigned or deactivated numbers from send lists, and numbers on state and national Do Not Call registries.
https://www.bloomreach.com/en/news

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature content from Barr Moses & Shane Murray, and Peter Lawrence.

Additional reading comes from Jay Graber, Kingsley Uyi Idehen, Stanford, and the National Science Foundation.

News comes from Expert[.]ai & Reveal Group, Databricks, Algolia, and Stilo.

Note that news items now link to the original source of the news rather than our 200 word summaries, which are always available here.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

Zero-ETL, ChatGPT, and the future of data engineering

“The post-modern data stack is coming. Are we ready?”.

Barr Moses & Shane Murray aim to help you prepare with this very readable post on promising components of future data stacks. (9 min)

https://towardsdatascience.com/zero-etl-chatgpt-and-the-future-of-data-engineering-71849642ad9c

Knowledge Graphs + Large Language Models = The ability for users to ask their own questions?

One area the article recommended above doesn’t address is the incorporation of knowledge graphs in data stacks. This is an area we are keenly interested in, especially in combination with LLMs. Peter Lawrence examines what may seem a less-than-obvious use-case (digital twins) for such a combination, and describes how a knowledge graph can be used to train an LLM to be significantly more valuable. He also illustrates why it is not easy. (12 min)

https://medium.com/@peter.lawrence_47665/knowledge-graphs-large-language-models-the-ability-for-users-to-ask-their-own-questions-e4afc348fa72

More Reading

• A marketplace for social Algorithmic choice, via Bluesky
• HowTo — Using CHAT-GPT to generate a Knowledge Graph from a Mermaid Entity via OpenLink Software Blog
• Just out… The Stanford 2023 AI Index Report, via Stanford
• Looking for proposals to help. Building the Prototype Open Knowledge Network (Proto-OKN) via National Science Foundation

All Gilbane Advisor issues

---

Content technology news

Expert[.]ai and Reveal Group partner to combine NLP and RPA

Expert[.]ai and Reveal Group partner to combine NLP and RPA — To help organizations extend the value in intelligent automation programs with natural language processing and understanding (NLP/NLU).
https://www.expert.ai/ ■ https://revealgroup.com/

Databricks announces Lakehouse for Manufacturing

Lakehouse for Manufacturing offers pre-built solutions, partner-designed Brickbuilder offerings and integrated AI capabilities.
https://www.databricks.com/solutions/industries/manufacturing-industry-solutions

Algolia introduces developer-friendly plan 

The AI Search and Discovery platform’s two new plans include a “Build” plan that is free and a “Grow” plan that offers low-priced scalability.
https://www.algolia.com

Stilo launches Migrate 5.0 

Migrate 5 provides automated content conversion from legacy content to structured DITA XML.
https://www.stilo.com/

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature content from OpenAI and Kurt Cagle.

Additional reading comes from Alex Hollender, Dean Allemang, and Tom Warren.

News comes from Adobe, Ontotext, Kobai, and Syncro Soft.

Note that news items now link to the original source of the news rather than our 200 word summaries, which are always available here.

The next issue will be published April 12. All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

ChatGPT plugins

Many, perhaps most, of the problems with large language models like GPT are due to data limitations (availability, age, quality). The ability to complement GPT data with external knowledge and data sources you own or trust, or are specific to your industry or business, increases model quality, context, and utility. LlamaIndex APIs can help you accomplish this. Now ChatGPT plugins take integration possibilities up a level. This post by the OpenAI team is the best place to get up-to-speed on the first batch of plugins, and what they’re planning. (12 min)

https://openai.com/blog/chatgpt-plugins

Nine ChatGPT tricks for knowledge graph workers

A helpful show-and-tell from Kurt Cagle… 

“Where things get interesting is in the realm of coding and semantics… most of these samples made use of ChatGPT+ for experimentation and the DaVince OpenAI Playground for generating more complex output.” (22 min)

https://thecaglereport.com/2023/03/16/nine-chatgpt-tricks-for-knowledge-graph-workers/

More Reading

• Design notes on the 2023 Wikipedia redesign via UX Collective
• LLM’s closing the KG gap via Dean Allemang
• You can now try Microsoft Loop, a Notion competitor with futuristic Office documents via The Verge

All Gilbane Advisor issues

---

Content technology news

Adobe announces multiple product updates

Adobe announcements focused on generative AI services, personalization, content management, product analytics, and content supply chain.
https://news.adobe.com/home/default.aspx

Ontotext releases Metadata Studio 3.2

Enables rapid text mining development based on an organization’s knowledge graph with non-technical contributors in the loop.
https://www.ontotext.com

Kobai launches Saturn Knowledge Graph 

Kobai Saturn platform users to query data at Lakehouse scale and integrate every use case and function into a single semantic layer.
https://www.kobai.io/products/kobai-saturn

Syncro Soft releases Oxygen XML suite 25.1 

Updates to Oxygen XML Editor, Author, Developer, Web Author, Publishing Engine, WebHelp, PDF Chemistry, and Scripting, and Oxygen Feedback.
https://www.sync.ro

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature content from Rama Ramakrishnan, and a course by, Anish Athalye, Curtis Northcutt, Jonas Mueller, Cody Coleman, Alexandra Zytek, & Sharon Zhou.

Additional reading comes from Benjamin Marie, Sarah Gooding, and Dmitry Kan.

News comes from Ontotext, Acquia, Lucid Software, and Databricks.

👉 No issue next week — we’ll be back March 29th.

Note that news items now link to the original source of the news rather than our 200 word summaries, which are always available here.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

The Road to ChatGPT — An informal explainer on how ChatGPT was built

Rama Ramakrishnan…
“I have written an informal “explainer” on how ChatGPT was built. I have tried to focus on the key ideas and have kept technical details to the bare minimum. Please view in full-screen mode. I hope you find it useful.”

This is excellent. You truly don’t need to be technical to follow it. Don’t let the 104 slides scare you off; you can fly through most of them. 

https://www.linkedin.com/feed/update/urn:li:activity:7038334518004482048/

Introduction to Data-Centric AI

“In real-world applications, data is messy and improving models is not the only way to get better performance. You can also improve the dataset itself rather than treating it as fixed. Data-Centric AI (DCAI) is an emerging science that studies techniques to improve datasets, which is often the best way to improve performance in practical ML applications.”

MIT has made this course freely available to all. Lectures and materials are online. This is a valuable resource for professional practitioners as well as students. (1 min for the course description – the rest is up to you).

https://dcai.csail.mit.edu

More Reading

• Traditional versus neural metrics for machine translation evaluation via Towards Data Science
• Automattic acquires ActivityPub plugin for WordPress, via WP Tavern
• Neural search frameworks: a head-to-head comparisonvia Dmitry Kan

All Gilbane Advisor issues

---

Content technology news

Ontotext releases GraphDB 10.2

Model Serving provides fully managed production machine learning (ML) capabilities natively within the Databricks Lakehouse Platform.
https://www.ontotext.com/

Acquia adds integrations to Digital Asset Management (DAM) System

The new integrations give customers more control over brand consistency, extend the value of content and data created in other systems.
https://www.acquia.com/

Lucid Software announces new integrations to enhance collaboration

Integrations streamline processes and workflows within a company’s tech stack and creates a foundation for effective and efficient collaboration.
https://lucid.co/

Databricks launches Databricks Model Serving

Model Serving provides fully managed production machine learning (ML) capabilities natively within the Databricks Lakehouse Platform.
https://www.databricks.com/

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature articles by Purvanshi Mehta, and Daniel Waisberg, Gaal Yahas, & Haim Daniel.

Additional reading from Jeremy Perdue, Rachel Gordon, and Nandan Grover.

News comes from Quark, TigerGraph, Slang Labs, and DeltaXML.

Note that news items now link to the original source of the news rather than our 200 word summaries, which are always available here.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

Indexing and Querying external KBs through GPT — GPT Index (LlamaIndex)

Purvanshi Mehta provides an overview of how you can use the GPT Index (now LlamaIndex) to customize large language models by connecting with external knowledge bases relevant to your application domain. She helpfully includes examples for different types of data sources. (6 min).

https://medium.datadriveninvestor.com/querying-external-kbs-through-gpt-gpt-index-llamaindex-fd8cbad2a4c

Bulk data export: a new and powerful way to access your Search Console data

From the Google Search Console Team…

You can configure an export in Search Console to get a daily data dump into your BigQuery project. The data includes all your performance data, apart from anonymized queries, which are filtered out for privacy reasons … This means you can explore your data to its maximum potential, joining it with other sources of data and using advanced analysis and visualization techniques. … This data export could be particularly helpful for large websites with tens of thousands of pages, or those receiving traffic from tens of thousands of queries a day (or both!). (3 min).

https://developers.google.com/search/blog/2023/02/bulk-data-export

More Reading

• Outlook Mac for all, free, no subscription, and native via Microsoft
• Large language models are biased. Can logic help save them? via MIT
• The ultimate review of GitHub Copilot for language translation via DataDrivenInvestor

All Gilbane Advisor issues

---

Content technology news

Quark releases Quark Publishing Platform NextGen v3.0

The content automation platform is designed to simplify the complexities associated with enterprise content lifecycle management.
https://www.quark.com ■ https://developer.quark.com

TigerGraph expands cloud capabilities

Provides a comprehensive, streamlined approach to deploy and maintain multiple graph database solutions with visual analytics and machine learning tools.
https://www.tigergraph.com/

Slang Labs launches CONVA

CONVA is a full-stack solution that provides smart and highly accurate multilingual voice search capabilities inside e-commerce apps.
https://www.slanglabs.in/media

DeltaXML eases HTML table comparison with XML Compare 14

The finer-grained detail for cells and rows are easier to display and provide greater understanding when reviewed.
https://www.deltaxml.com

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact

This week we feature articles by Jacob Steinhardt, and Dean Allemang.

Additional reading from Rocío Txabarriaga, Eric Broda, and Tony Seale.

News comes from MadCap Software & IXIASOFT, BetterCommerce, Wondershare, and Contentstack.

Note that news items now link to the original source of the news rather than our 200 word summaries, which are always available here.

All previous issues are available at https://gilbane.com/gilbane-advisor-index/

---

Opinion / Analysis

Emergent deception and emergent optimization

Emergent properties are in common in nature, and are often surprising. They are also found in machine learning. Jacob Steinhardt has a series of posts on emergence in machine learning worth checking out, but you can start with his most recent, and timely, piece. (17 min).

I’ve previously argued that machine learning systems often exhibit emergent capabilities, and that these capabilities could lead to unintended negative consequences. But how can we reason concretely about these consequences? … I’ll describe two specific emergent capabilities that I’m particularly worried about: deception (fooling human supervisors rather than doing the intended task), and optimization (choosing from a diverse space of actions based on their long-term consequences).

https://bounded-regret.ghost.io/emergent-deception-optimization/

Why I’m not excited about RDF-Star

Well, the title is a bit clickbaity. But Dean Allemang’s article illustrates an important point about RDF modeling in general. And if like me, you weren’t aware of RDF-Star, an added benefit is you’ll learn enough to consider how you might use it when the W3C standard becomes a recommendation. (10 min).

https://medium.com/@dallemang/why-im-not-excited-about-rdf-star-5f1993fd0ead

More Reading

• Building your data network via Tony Seale
• Managing a federated data product ecosystem, via Towards Data Science
• How large language models can get better at machine translation of rare words via Slator

All Gilbane Advisor issues

---

Content technology news

MadCap Software acquires IXIASOFT

Adds enterprise DITA CCMS to support content strategies for creating, translating, and delivering consistent, up-to-date content tailored to roles.
https://www.madcapsoftware.com ■ https://www.ixiasoft.com

Contentstack announces Contentstack Launch

Extends Contentstack’s product suite, providing a composable, automated, digital experience stack from the front-end to the back-end.
https://www.contentstack.com/

BetterCommerce adds headless CMS functionality to its commerce stack

The headless, composable CMS functionality joins existing modules in the commerce stack including PIM, eCommerce, OMS, Analytics and Engage.
https://www.bettercommerce.io

Wondershare releases EdrawMind 10.5

Features new collaborative mind mapping and brainstorming tools to design solutions collaboratively and respond to trends and changes.
https://www.edrawsoft.com/edrawmind/

All content technology news

---

The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, web, data, and digital experience technology and information professionals. We publish recommended articles and content technology news weekly. We do not sell or share personal data.

Subscribe | View online | Editorial policy | Privacy policy | Contact